- Email: [email protected]
The real number of road traffic accident casualties in the Netherlands: A year-long survey
Accid. And. & Prev. Vol. 22, No. 4. pp. 371-378. Printed in Great Britain.
1990
OcKll-4575KxJ s3.00+ .cm 0 1990 Pergamon Press plc
THE REAL NUMBER OF ROAD TRAFFIC ACCIDENT CASUALTIES IN THE NETHERLANDS: A YEAR-LONG SURVEY STEPHEN HARRIS SWOV Institute for Road Safety Research, P.O. Box 170, 2260 Leidschendam, The Netherlands. (Received 11 Ocrober 1989) Abstract-Between August 1986 and Juty 1987 more than 24,000 households, containing nearly 67,000 persons, were surveyed by telephone about traffic injuries during the past three months, Expressed on an annual basis, approximately 430,000 people, or about 1 in 34 of the Dutch population, had suffered some sort of iniury in a road accident. The road traffic morbidity was, therefore, 2,942 per 100,000 inhabitants: df these, about 135,000 had to be treated in hospital (20.000 as inoatients). More than 100,000 did not need treatment. Cyclists formed by far the Iargest category of road user, but mopedists had the highest injury rate per kilometer travelled. 210,000 of these casualties fell within the definition for recording by the police. The police recorded only 49,748 traffic casualties, or about 25%, during the same period. The police data were not representative; the completeness declined according to severity of the injuries: inpatients, about 70%; outpatients 26%; extramural about 11%. Cyclists (ll%), children (9%), and single vehicle accidents (5%) were very much underrepresented. The largest category of road user is cyclists, not car occupants as indicated by the police data. A number of recommendations are made for supplementing the police data and the existing hospital inpatient data. These include extending the Home Accident Recording System of outpatients and the General Practitioner Panel to include road accident victims. Together a representative sample of 95% of all those receiving medical treatment would thus be obtained.
INTRODUCTION
We road safety researchers do not know the size of our own field of study. We do not even know the number of injury accidents or the number of those injured. It has however long been suspected that in the Netherlands, as in other countries, the police recording of road traffic injury accidents and victims, which provide the official accident statistics, is incomplete and probably not representative. Past research in the Netherlands (Maas and Harris 1984) had shown that the completeness of police records for inpatients was approximately 85% and stable during the latter half of the 1970s (it has since declined to slightly less than 70% in 1988!) and was not completely representative for either age or modal split: children and cyclists were especially underreported. As fatal accident records, in comparison with death certificate data, were not shown to be less than 100% complete, it was assumed that the completeness and representativity declines as injury severity declines. This article describes a survey that was carried out in the Netherlands between August 1986 and July 1987. This survey was part of a combined project with the Institute for Consumer Safety and the University of Limburg. It was financed by the Ministries of Transport and Health. Questions were asked about all types of accidents: traffic, domestic, leisure, sport, and industrial. It was published in a combined report in 1988 (van Montfoort et al.). Since then SWOV has published (so far only in Dutch) a separate, detailed report on the traffic casualties (Harris 1989), on which this article is based. OBJECTIVES
The objectives of this survey were twofold. The primary objective was to determine the “actual” or real number of casualties: actual in this case is defined as that number reported by those involved; the survey method was used, as there were no complete data. The second objective was to establish the completeness and representativity of the 371
372
S. HARRIS
police data, especially in relation to injury severity. This article pays special attention to the second objective, as this is presumed to be or more interest internationally. DEFINITIONS
In the survey questionnaire, a road traffic accident was defined as “an accident as driver or passenger of a vehicle or as a pedestrian.” This was deliberately broader than the international definition of the Convention on Road Traffic (held in Vienna in 1968), which limits accidents to those “which occurred or originated on a way or street open to public traffic, which resulted in one or more persons being killed or injured, and in which at least one moving vehicle was involved” (United Nations 1989). We wanted to know how many accidents occur that fall outside the definition and, therefore, by definition, outside the police registration. By asking whether the accident had occurred on a public road and what types of road users were involved in the accident, it was possible to compare the survey results with police data. Furthermore, we did not wish for respondents to have to make a selection themselves in case they should think that not all accidents were important and, therefore, exclude other accidents that did not fall within the definition. Znjury was deliberately not defined in the questionnaire. This was also to avoid the situation of the respondent making a selection by thinking that certain “(s)light” injuries, such as scratches and bruises, were not important enough to be mentioned. Injury is not defined by the international definition, which speaks only of “serious or slight injuries.” Questions about any hospital treatment made it possible to compare the survey results with police data. Injuries could not be classified according to the International Classification of Diseases (9th revision) of the World Health Organization (1977), because the source of the information (the victims themselves) were not medically trained. The existing classification of the continuous Home and Leisure Accident Surveillance System of the Consumer Safety Institute by 25 types of injury and 40 injury locations, was used. Injury severity, therefore, could not be expressed by means of, for example, the Abbreviated Injury Scale (AIS) or Maximum AIS. An ordinal scale was developed using the “most important treatment” that the victim underwent. The order of importance in descending order of severity was: inpatient, outpatient, (own) doctor, other professional treatment, semiprofessional treatment, nonprofessionaland self-help, no treatment. The last five are totalized as “not in hospital” because the police data make no distinction between them. METHOD
During the survey period from August 1986 to July 1987, 24,141 randomly selected households containing 66,804 persons were interviewed by telephone by means of Computer Assisted Telephone Interviewing (CATI). Ninety-three percent of Dutch households have a telephone. The male or female head of the household was asked which, if any, members of the household had had a road accident resulting in injury, however light, during the previous three months. If there had been an accident, the interview was continued with the victim(s) unless this was a child ten years old or younger. In this case, the parent answered the further questions about the accident and the injuries. These questions were, of course, limited to what one could reasonably expect them to know and from which one could reasonably expect an honest answer. Therefore, there were no questions, for example, on driving speed or who was to blame. The three-month’ recall period was chosen because a previous, unpublished SWOV survey in 1976 on accidents during the whole of 1975 had shown that victims of road accidents had a stable recall rate up to three months. After this, i.e. longer than “three months ago,” the recall rate dropped rapidly. The nonresponse was 16%, resulting in a sample that was geographically not quite representative and an underrepresentation of single-person households. The distribution
373
Real number of road traffic accident casualties
of road traffic injuries did not, however, household size did vary somewhat.
vary by geographical
region; distribution
by
RESULTS
All results should be regarded as minimum values because of the use of the survey method. The survey method can lead to memory loss especially if the accident was not serious, and concealment, especially if feelings of guilt are involved. The extent of this possible underreporting is unknown. Furthermore, the results refer to accidents of residents of the Netherlands and not to all accidents on Dutch roads. Only the household population was interviewed; not the 2% institutionalized population such as those living in old peoples’ homes. Overreporting as a result of “telescoping” (reporting an accident that occurred earlier than three months ago because it seemed more recent) is minimal because nearly everyone was able to provide the exact date of the accident. Deliberate lying, however, cannot be entirely discounted. Four hundred ninety-one respondents said that they had been injured in a road accident during the previous three months. Weighted for the sample fraction and the recall period, this meant 430,~ casualties, with a margin of 35,000 (P = 0.05), during the la-month period. This is 1 in 34 Dutch population of 14.615 million (as of January 1, 1987) or a road traffic morbidity of 2,942 per 100,000 inhabitants. 18,000 victims were admitted as inpatients in a hospital. This figure, considering the sample size, compares well with the hospitals’ figure of approximately 21,000. A further 117,000 were treated as outpatients, a figure that was unknown until now as there is no national recording system. Hospital treatment was, therefore, needed by 135,000 casualties. Another 105,000 were treated by their doctor without needing hospital treatment. Slightly more than 100,000 received no treatment at all. Cyclists formed by far the largest category of road user, with 211,000 casualties; a fact that confirms Holland’s reputation as a country of bicycles. They were followed by pedestrians (73,~0), car occupants (69,0~), mopedists (60,~), motorcyclists (lO,~), and 7,000 others. The category with the highest injury rate per milliard (1,000 million) kilometers travelled (using data from the Central Bureau of Statistics National Travel Survey) were the mopedists with a rate almost twice as high (33,333) as that of the second highest category, the cyclists (18,347). Motorcyclists, who have the highest death rate, were not much further behind (16,667); pedestrians (14,038) were next. Car occupants had by far the lowest rate of only 590, with 702 for drivers and only 432 for passengers. Car occupants, however, had the highest proportion of inpatients and cyclists the lowest. Looking at the combinations of road-user categories involved, the largest combination was accidents involving two vehicles. Accidents with pedestrians and objects were also important. Also worth mentioning are the 25,000 pedestrians injured by tripping and slipping-accidents that are by definition excluded from the police recording system because no moving vehicle was involved (see DEFINITIONS). As far as age is concerned, by far the highest rate (per milliard kilometers) was found among 15 17-year-olds (9,107), more than half of whom were mopedists, followed by children up to the age of 14 (6,707), of whom 80% were cyciists, and the 18- and 19-year-olds (4,737), of whom half were mopedists. Among adults there was little difference (approximately 2000) except among those 65 and older, whose rate was 3,025 and of whom half were pedestrians. As many women were injured as men, but their injury rate was about one-and-ahalf times as high. About 70% of the injuries occurred in accidents inside built-up areas where the accident rate was eight times that outside built-up areas. Only 4% took place on motorways, and their accident rate was only one tenth that of the other roads. The highest concentration was found between 2 P.M. and 4 P.M. with all categories of road user having more or less the same diurnal pattern. By far the highest injury rate, however, was found between 12 A.M. and 4 A.M..
374
S. HARRIS Table 1. From injury to police registration All casualties reported in survey 0.w. fulfil registration requirements 0.w. police present O.W. registered by police
430,000 210,000 95,000 49,748
100% 49%
100% 45% 24%
100% 52%
O.W. = of which
On average about one and one-half injuries per victim were sustained. The most common injuries (40%) were bruises and scrapes to the arms and legs and cuts and scrapes on the head (15%) The highest proportion of injuries treated in hospital were head injuries: half of the inpatients had head injuries. Car occupants had a higher proportion of head injuries than other categories of road user. Arm and, especially, leg injuries were the most common with cyclists, mopedists, and pedestrians. THE
COMPARISON
WITH
POLICE
DATA
For testing the completeness and representativity of the police data, only those casualties are taken into consideration who comply with the definition for inclusion. These are: victims of accidents-on the public highway, involving at least one moving vehicle (the Netherlands uses the international definition)-whose injuries are worse than “very light.” Very light injuries are excluded from Dutch police statistics, but are not defined-the example of a “scratch” is simply given on the back of the report form. For this comparison very light injuries were considered to be those not requiring any professional treatment. Using these three exclusions 210,000 of the 430,000 casualties remained to be compared with the police data. Of these, the police were present in 95,000 cases according to the respondents (see Table 1). During the survey period the police recorded 49,748 people nonfatally injured. This was only 52% of the 95,000 cases in which they had been present and only 24% of the survey total of 210,000 and thus very incomplete. It will be noticed that there seems to be a pattern. Each successive number of injured is approximately half of the preceding number: 1. half the total number of casualties should be recorded; 2. the police are present for half of the cases they should record; 3. the police record half of the cases where they are present. This pattern is probably coincidental because the processes involved between the various steps are different. The police have no control over the total number of accidents; little control over whether in reality an accident falls within the definition (they can, however, decide which injuries are not serious enough); some control over whether, when called, they go to the scene; but total control over whether an accident is reported or not. For the Netherlands, however, this “halving” pattern would seem to be a useful rule of thumb for the time being. The completeness of approximately only 1 in 4 is much worse than found in the United States NASS survey in 1979 (Greenblatt et al. 1981). Using a similar method Table 2. Completeness Treatment
n
Casualties
-
by most important treatment Police
Completeness
Factor
(100%) 79% 26% 11%
(1) 1.3 3.8 8.8
24%
4.2
(Deaths) Inpatient Outpatient Not in hospital
21 106 113
18,000 93,000 99,000
( 1,492) 14,262 24,232” 11,254t
Total
240
210,000
49,748
(x’ = 97.03; df = 2; P = (0.01) Factor = Casualties/Police. * = taken to hospital but not admitted, or admission unknown t = not taken to hospital, or unknown
Real number of road traffic accident casualties Table 3. Comparison of the sub-registrations
375
by most important treatment Within police definition Present
Most important treatment
All survey casualties
Outside police definition
Inpatient Outpatient No hospital
18,000 117,000 292,000
0 24,000 193,000
18,000 93,000 99,000
Total
430,000
220,000
210,000
Not present
Total
Total
No rec.
Recorded
3,000 48,000 64,000
15,000 45,000 35,000
738 20,768 23,746
14,262 24,232* 11,254t
115,OOu
95,000
45,252
49,748
* = taken to hospital but not admitted t = not taken to hospital, or unknown
they found only approximately 0.3 unrecorded injury accidents for every 1 recorded injury accident. The police data were also far from representative. They recorded 79% of all inpatients (actually only 70% if compared with the approximately 20,000 from the National Medical Registration instead of from the survey), but only 26% of all outpatients and only 11% of those not needing hospital treatment (Table 2, see also Fig. 1). It is assumed that all victims who, according to the police, were taken to hospital and either were not admitted or whose admission is not known, were treated by the hospital as outpatients. To complete the picture, when compared with death certificate data, the police data on road accident deaths have not been shown to be less than 100%. Their coverage, therefore, declines rapidly with injury severity. This is in agreement with the findings of Tunbridge and Everest (1988). If the police were to record every casualty where they had been present, their data would still not be representative for injury severity. It would contain 83% of all inpatients, 48% of all outpatients, and 35% of all those not treated in a hospital (Table 3). (In addition, the completeness of data for Material Damage Only accidents is probably considerably lower than the 11% of those not treated in hospital.) Very much underrepresented were cyclists (11%) (Table 4, see also Fig. 2), especially, children on bikes (5%), and single vehicle bicycle accidents (2%). Single vehicle accidents in general (in accordance with Tunbridge and Everest 1988), and especially those in which not even an object was hit, so-called “single-sided accidents” in Dutch, (5%) were much underrepresented (9%), (Table 5, see also Fig. 3). The same applied to children in general (9%). All these results reflect also the incompleteness of police data in comparison with hospital inpatient data (Maas and Harris 1984). Relatively overrepresented, i.e. considerably more than the average of 24%, were car occupants (41%), especially those between 18 and 29 years old (54%). This also applied to accidents between two vehicles of which one was a car (40%) or one a moped (54%) and to accidents occurring at night between the hours of 10 P.M. and 7 A.M. This last result was unexpected because we had thought at night, with less police on duty and more crime, reporting would be less complete than during the daytime. The largest category of road user as the number of injured are concerned are not car occupants, which is what the police data indicates, but cyclists-they make up half Table 4. Completeness Mode of transport
n
Car Bicycle Moped Motorcycle Pedestrian Rest
50 122 41 6 18 3
Total
240
Casualties
by mode of transport Police
Completeness
Factor
44,000 106,000 36,000 5,0@3 16,000 3,000
18,269 11,734 11,998 1,961 4,076 1,710
41% 11% 33% 39% 25% 57%
2.4 9.0 3.0 2.5 3.9 1.8
210,000
49,748
24%
4.2
(xl = 100.86; df = 5; P < 0.01)
S.HARRIS
376
Table 5. Completeness
by collision type
Collision type
n
Casualties
Police
Completeness
Factor
with: Vehicle Object Nothing else Unknown
130 30 77 3
114,000 26,000 67,000 -
40,055 5,523 3,258 912
35% 21% 5% -
2.9 8.0 21 -
Total
240
210,000
49,748
24%
4.2
(x’ = 252.83; df = 3; P < 0.01)
of all wounded; not twens (people in their twenties), but children up to the age of fourteen; no mopedists between 15 and 29 years old, but child cyclists; not cars in collision with another vehicle, but single-vehicle cycle accidents. The ranking in the casualty rate (per milliard kilometers travelled) by category of road user does not alter much if the survey results are used instead of the police data. As far as age group is concerned the ranking does change; according to the survey data, children up to the age of 14 years become much more important in comparison with the police data. Women, not men, have the higher casualty rate. The early-evening and lunch hours become much more important, and the late-evening and nighttime hours less so. The ratios between weekdays and weekend, and inside and outside built-up areas change little. The two factors of seemingly greatest importance for determining whether an injury accident will be recorded are: injury severity and involvement of a motor vehicle. RECOMMENDATIONS
Following are recommendations road traffic accidents:
for improving
the recording
of those injured in
1. The public, those involved in the accident as well as bystanders, should report more injury accidents to the police. 2. The police should report all injury accidents conforming to the definition instead of only half of those where they were present. The expectations are low as far as improving the police recording in the future is concerned. Further computerization at the scene of accidents and at police stations, to decrease the amount of time needed to record each accident, does not guarantee the recording of more accidents. This is confirmed by certain police authorities here. Extra manpower for accident recording is certainly not readily available and may never be. There are greater hopes for the already intensively used integral inpatient data of the National Medical Registration as complementary to police data. Since 1987 all hospitals participate in this registration. 100 j
%
100
80
70% irr 60 P. a I 40 _ i e c 20 - n t 8
0
I
0
-
80
-
60
-
40
-
20
26% 11%
out-patients
notinhospital I
40
I
I
80
I
I
I
120
Fig. 1. Completeness
I
160
by treatment.
I
I'
200
0 x1000
Real number of road traffic accident casualties
60 40 20 0 210
0 Fig. 2. Completeness
by mode of transport.
3. The main recommendation is to obtain more nearly complete outpatient data by adding road traffic accident victims to the existing continuous Home and Leisure Accident Surveillance System of the Consumer Safety Institute; a representative 10% sample of 14 hospital Accident and Emergency Departments. Since the original draft report was presented to the Ministry of Transport it has commissioned SWOV to carry out a trial registration. This ran in two hospitals (one a teaching hospital, the other a general hospital) during the months of September, October, and November 1989. If successful (i.e. the necessary details can be obtained), the addition of road accident victims will become continuous. 4. Further recommended is exploration of the possibility of recording road victims by general practitioners, using the existing continuous General Practitioner Panel; a panel of 60 General Practitioners whose 45 practices cover 1% of the population. 5. Including road accident victims in these two registrations (outpatients and General Practitional patients), in combination with the existing inpatient data, would result in a continuous recording of practically all those receiving professional medical treatment (95% according to this survey). 6. The possible establishment of a separate Road Accident Recording Corps is also recommended. This would complement the police rather in the same way as Parking Wardens do. 7. Also recommended is the possibility of making this present one-year survey continuous, or at least repeating it, for example, every five years. %
%
100
100 80 -
-
80
60 -
-
60
35%
-
40
vehide
-
20
40 -
..,. 20 -
0
.’ , 0
I 40
I
‘. I
I
80
I
120
I
I
160
I
I 200
0 x1000
NB: the width of each staff is proportional to the number of survey injured. Fig. 3. Completeness
by collision type.
378
S. HARRIS
It must, however, be emphasized that the additional registrations recommended are primarily of victims and not accidents. They do not, therefore, contain, and probably never will do, many of the precrash and crash variables available in the police data. There is, however, the theoretical possibility of follow-up interviews with patients. These would have to take place soon after the accident and there are considerable privacy problems in getting patients’ names and addresses. As long as this is the case the police data remain indispensable for road safety research. INTERNATrONAL
IMPLrCATIONS
The extent of incompleteness of records, especially of outpatients, the extra-murally treated, and accidents not involving a motor vehicle, will not be the same in every country and will probably vary greatly. For example, the number of police-reported casualties for every reported fatality varied by more than a factor 4 between developed countries in 1987 (United Nations 1989)! In Denmark and Ireland there were only 17 casualties per fatality, the highest in Europe was 60 for Great Britain (in the Netherlands 33), and in Japan there were 77. It is, however, clear from these results for the Netherlands that the situation in other developed countries, whether better or worse than in the Netherlands, will also be cause for concern. In the United States the incompleteness of records is apparently much less- only 30% ~Greenblatt et al. 1981). Similar surveys in other countries will undoubtedly show to what extent reliance on the official (police) accident data can lead to inappropriate priorities. They will also show, therefore, that complementary data, as recommended by Hauer and Hakkert (1987) is needed. In our opinion, to a great extent this should be hospital data that include outpatients as well as inpatients. At the moment international comparisons of road safety should be limited to data on fatal accidents and fatalities. Use of police data on injury accidents and casualties should be avoided because of their great incompleteness and lack of representativity. They are not a valid indicator of road safety-in fact, they are misleading. Use of only those (nonfatal) victims being admitted to hospital (inpatients) may be valid if (a) the criteria for admission, and (b) the completeness of the police data are approximately the same in most countries; or at least known from year to year ~OE~~/SWOV 1988).
REFERENCES Greenblatt, J.; et al. National Accident Sampling System nonreported accident survey. Washington: National Highway Traffic Administration; 1981. Harris, S. Verkeersgewonden geteld en gemeten (Traffic casualties counted and measured)-with English summary. (R-89-13), Leidschendam: SWOV; 1989. Hauer, E.; Hakkert, A.S. The extent and some implications of incomplete accident reporting. 66th TRB meeting. Washington, DC: Transportation Research Board; 1987. Maas, M.W.; Harris, S. Police recording of road accident inpatients, Accid. Anal. Prev. 16: 167-184; 1984. Montfoort, G.L.M. van; Galen, dr. W.Ch.C. van; Harris, S. Ongevallen in Nederland (Accidents in the Netherlands)-with English summary. Amsterdam: Stichting Consument en Veiligheid; 1988. OECDISWOV. Framework for consistent traffic and accident statistical data bases. Leidschendam: SWOV; 1988. Tunbridge, R.J.; Everest, J.T. An assessment of the under reporting of road accident casualties in relation to injury severity. IRCOBI Conference, Bergisch Gladbach (FRG), September, 1988. United Nations. Statistics of road traffic accidents 1987, data and annex 1: Definitions and general notes. New York: UN; 1989. WHO. Manual of the international statistical classification of diseases, injuries and causes of death. Geneva: World Health Organization; 1977.
1990
OcKll-4575KxJ s3.00+ .cm 0 1990 Pergamon Press plc
THE REAL NUMBER OF ROAD TRAFFIC ACCIDENT CASUALTIES IN THE NETHERLANDS: A YEAR-LONG SURVEY STEPHEN HARRIS SWOV Institute for Road Safety Research, P.O. Box 170, 2260 Leidschendam, The Netherlands. (Received 11 Ocrober 1989) Abstract-Between August 1986 and Juty 1987 more than 24,000 households, containing nearly 67,000 persons, were surveyed by telephone about traffic injuries during the past three months, Expressed on an annual basis, approximately 430,000 people, or about 1 in 34 of the Dutch population, had suffered some sort of iniury in a road accident. The road traffic morbidity was, therefore, 2,942 per 100,000 inhabitants: df these, about 135,000 had to be treated in hospital (20.000 as inoatients). More than 100,000 did not need treatment. Cyclists formed by far the Iargest category of road user, but mopedists had the highest injury rate per kilometer travelled. 210,000 of these casualties fell within the definition for recording by the police. The police recorded only 49,748 traffic casualties, or about 25%, during the same period. The police data were not representative; the completeness declined according to severity of the injuries: inpatients, about 70%; outpatients 26%; extramural about 11%. Cyclists (ll%), children (9%), and single vehicle accidents (5%) were very much underrepresented. The largest category of road user is cyclists, not car occupants as indicated by the police data. A number of recommendations are made for supplementing the police data and the existing hospital inpatient data. These include extending the Home Accident Recording System of outpatients and the General Practitioner Panel to include road accident victims. Together a representative sample of 95% of all those receiving medical treatment would thus be obtained.
INTRODUCTION
We road safety researchers do not know the size of our own field of study. We do not even know the number of injury accidents or the number of those injured. It has however long been suspected that in the Netherlands, as in other countries, the police recording of road traffic injury accidents and victims, which provide the official accident statistics, is incomplete and probably not representative. Past research in the Netherlands (Maas and Harris 1984) had shown that the completeness of police records for inpatients was approximately 85% and stable during the latter half of the 1970s (it has since declined to slightly less than 70% in 1988!) and was not completely representative for either age or modal split: children and cyclists were especially underreported. As fatal accident records, in comparison with death certificate data, were not shown to be less than 100% complete, it was assumed that the completeness and representativity declines as injury severity declines. This article describes a survey that was carried out in the Netherlands between August 1986 and July 1987. This survey was part of a combined project with the Institute for Consumer Safety and the University of Limburg. It was financed by the Ministries of Transport and Health. Questions were asked about all types of accidents: traffic, domestic, leisure, sport, and industrial. It was published in a combined report in 1988 (van Montfoort et al.). Since then SWOV has published (so far only in Dutch) a separate, detailed report on the traffic casualties (Harris 1989), on which this article is based. OBJECTIVES
The objectives of this survey were twofold. The primary objective was to determine the “actual” or real number of casualties: actual in this case is defined as that number reported by those involved; the survey method was used, as there were no complete data. The second objective was to establish the completeness and representativity of the 371
372
S. HARRIS
police data, especially in relation to injury severity. This article pays special attention to the second objective, as this is presumed to be or more interest internationally. DEFINITIONS
In the survey questionnaire, a road traffic accident was defined as “an accident as driver or passenger of a vehicle or as a pedestrian.” This was deliberately broader than the international definition of the Convention on Road Traffic (held in Vienna in 1968), which limits accidents to those “which occurred or originated on a way or street open to public traffic, which resulted in one or more persons being killed or injured, and in which at least one moving vehicle was involved” (United Nations 1989). We wanted to know how many accidents occur that fall outside the definition and, therefore, by definition, outside the police registration. By asking whether the accident had occurred on a public road and what types of road users were involved in the accident, it was possible to compare the survey results with police data. Furthermore, we did not wish for respondents to have to make a selection themselves in case they should think that not all accidents were important and, therefore, exclude other accidents that did not fall within the definition. Znjury was deliberately not defined in the questionnaire. This was also to avoid the situation of the respondent making a selection by thinking that certain “(s)light” injuries, such as scratches and bruises, were not important enough to be mentioned. Injury is not defined by the international definition, which speaks only of “serious or slight injuries.” Questions about any hospital treatment made it possible to compare the survey results with police data. Injuries could not be classified according to the International Classification of Diseases (9th revision) of the World Health Organization (1977), because the source of the information (the victims themselves) were not medically trained. The existing classification of the continuous Home and Leisure Accident Surveillance System of the Consumer Safety Institute by 25 types of injury and 40 injury locations, was used. Injury severity, therefore, could not be expressed by means of, for example, the Abbreviated Injury Scale (AIS) or Maximum AIS. An ordinal scale was developed using the “most important treatment” that the victim underwent. The order of importance in descending order of severity was: inpatient, outpatient, (own) doctor, other professional treatment, semiprofessional treatment, nonprofessionaland self-help, no treatment. The last five are totalized as “not in hospital” because the police data make no distinction between them. METHOD
During the survey period from August 1986 to July 1987, 24,141 randomly selected households containing 66,804 persons were interviewed by telephone by means of Computer Assisted Telephone Interviewing (CATI). Ninety-three percent of Dutch households have a telephone. The male or female head of the household was asked which, if any, members of the household had had a road accident resulting in injury, however light, during the previous three months. If there had been an accident, the interview was continued with the victim(s) unless this was a child ten years old or younger. In this case, the parent answered the further questions about the accident and the injuries. These questions were, of course, limited to what one could reasonably expect them to know and from which one could reasonably expect an honest answer. Therefore, there were no questions, for example, on driving speed or who was to blame. The three-month’ recall period was chosen because a previous, unpublished SWOV survey in 1976 on accidents during the whole of 1975 had shown that victims of road accidents had a stable recall rate up to three months. After this, i.e. longer than “three months ago,” the recall rate dropped rapidly. The nonresponse was 16%, resulting in a sample that was geographically not quite representative and an underrepresentation of single-person households. The distribution
373
Real number of road traffic accident casualties
of road traffic injuries did not, however, household size did vary somewhat.
vary by geographical
region; distribution
by
RESULTS
All results should be regarded as minimum values because of the use of the survey method. The survey method can lead to memory loss especially if the accident was not serious, and concealment, especially if feelings of guilt are involved. The extent of this possible underreporting is unknown. Furthermore, the results refer to accidents of residents of the Netherlands and not to all accidents on Dutch roads. Only the household population was interviewed; not the 2% institutionalized population such as those living in old peoples’ homes. Overreporting as a result of “telescoping” (reporting an accident that occurred earlier than three months ago because it seemed more recent) is minimal because nearly everyone was able to provide the exact date of the accident. Deliberate lying, however, cannot be entirely discounted. Four hundred ninety-one respondents said that they had been injured in a road accident during the previous three months. Weighted for the sample fraction and the recall period, this meant 430,~ casualties, with a margin of 35,000 (P = 0.05), during the la-month period. This is 1 in 34 Dutch population of 14.615 million (as of January 1, 1987) or a road traffic morbidity of 2,942 per 100,000 inhabitants. 18,000 victims were admitted as inpatients in a hospital. This figure, considering the sample size, compares well with the hospitals’ figure of approximately 21,000. A further 117,000 were treated as outpatients, a figure that was unknown until now as there is no national recording system. Hospital treatment was, therefore, needed by 135,000 casualties. Another 105,000 were treated by their doctor without needing hospital treatment. Slightly more than 100,000 received no treatment at all. Cyclists formed by far the largest category of road user, with 211,000 casualties; a fact that confirms Holland’s reputation as a country of bicycles. They were followed by pedestrians (73,~0), car occupants (69,0~), mopedists (60,~), motorcyclists (lO,~), and 7,000 others. The category with the highest injury rate per milliard (1,000 million) kilometers travelled (using data from the Central Bureau of Statistics National Travel Survey) were the mopedists with a rate almost twice as high (33,333) as that of the second highest category, the cyclists (18,347). Motorcyclists, who have the highest death rate, were not much further behind (16,667); pedestrians (14,038) were next. Car occupants had by far the lowest rate of only 590, with 702 for drivers and only 432 for passengers. Car occupants, however, had the highest proportion of inpatients and cyclists the lowest. Looking at the combinations of road-user categories involved, the largest combination was accidents involving two vehicles. Accidents with pedestrians and objects were also important. Also worth mentioning are the 25,000 pedestrians injured by tripping and slipping-accidents that are by definition excluded from the police recording system because no moving vehicle was involved (see DEFINITIONS). As far as age is concerned, by far the highest rate (per milliard kilometers) was found among 15 17-year-olds (9,107), more than half of whom were mopedists, followed by children up to the age of 14 (6,707), of whom 80% were cyciists, and the 18- and 19-year-olds (4,737), of whom half were mopedists. Among adults there was little difference (approximately 2000) except among those 65 and older, whose rate was 3,025 and of whom half were pedestrians. As many women were injured as men, but their injury rate was about one-and-ahalf times as high. About 70% of the injuries occurred in accidents inside built-up areas where the accident rate was eight times that outside built-up areas. Only 4% took place on motorways, and their accident rate was only one tenth that of the other roads. The highest concentration was found between 2 P.M. and 4 P.M. with all categories of road user having more or less the same diurnal pattern. By far the highest injury rate, however, was found between 12 A.M. and 4 A.M..
374
S. HARRIS Table 1. From injury to police registration All casualties reported in survey 0.w. fulfil registration requirements 0.w. police present O.W. registered by police
430,000 210,000 95,000 49,748
100% 49%
100% 45% 24%
100% 52%
O.W. = of which
On average about one and one-half injuries per victim were sustained. The most common injuries (40%) were bruises and scrapes to the arms and legs and cuts and scrapes on the head (15%) The highest proportion of injuries treated in hospital were head injuries: half of the inpatients had head injuries. Car occupants had a higher proportion of head injuries than other categories of road user. Arm and, especially, leg injuries were the most common with cyclists, mopedists, and pedestrians. THE
COMPARISON
WITH
POLICE
DATA
For testing the completeness and representativity of the police data, only those casualties are taken into consideration who comply with the definition for inclusion. These are: victims of accidents-on the public highway, involving at least one moving vehicle (the Netherlands uses the international definition)-whose injuries are worse than “very light.” Very light injuries are excluded from Dutch police statistics, but are not defined-the example of a “scratch” is simply given on the back of the report form. For this comparison very light injuries were considered to be those not requiring any professional treatment. Using these three exclusions 210,000 of the 430,000 casualties remained to be compared with the police data. Of these, the police were present in 95,000 cases according to the respondents (see Table 1). During the survey period the police recorded 49,748 people nonfatally injured. This was only 52% of the 95,000 cases in which they had been present and only 24% of the survey total of 210,000 and thus very incomplete. It will be noticed that there seems to be a pattern. Each successive number of injured is approximately half of the preceding number: 1. half the total number of casualties should be recorded; 2. the police are present for half of the cases they should record; 3. the police record half of the cases where they are present. This pattern is probably coincidental because the processes involved between the various steps are different. The police have no control over the total number of accidents; little control over whether in reality an accident falls within the definition (they can, however, decide which injuries are not serious enough); some control over whether, when called, they go to the scene; but total control over whether an accident is reported or not. For the Netherlands, however, this “halving” pattern would seem to be a useful rule of thumb for the time being. The completeness of approximately only 1 in 4 is much worse than found in the United States NASS survey in 1979 (Greenblatt et al. 1981). Using a similar method Table 2. Completeness Treatment
n
Casualties
-
by most important treatment Police
Completeness
Factor
(100%) 79% 26% 11%
(1) 1.3 3.8 8.8
24%
4.2
(Deaths) Inpatient Outpatient Not in hospital
21 106 113
18,000 93,000 99,000
( 1,492) 14,262 24,232” 11,254t
Total
240
210,000
49,748
(x’ = 97.03; df = 2; P = (0.01) Factor = Casualties/Police. * = taken to hospital but not admitted, or admission unknown t = not taken to hospital, or unknown
Real number of road traffic accident casualties Table 3. Comparison of the sub-registrations
375
by most important treatment Within police definition Present
Most important treatment
All survey casualties
Outside police definition
Inpatient Outpatient No hospital
18,000 117,000 292,000
0 24,000 193,000
18,000 93,000 99,000
Total
430,000
220,000
210,000
Not present
Total
Total
No rec.
Recorded
3,000 48,000 64,000
15,000 45,000 35,000
738 20,768 23,746
14,262 24,232* 11,254t
115,OOu
95,000
45,252
49,748
* = taken to hospital but not admitted t = not taken to hospital, or unknown
they found only approximately 0.3 unrecorded injury accidents for every 1 recorded injury accident. The police data were also far from representative. They recorded 79% of all inpatients (actually only 70% if compared with the approximately 20,000 from the National Medical Registration instead of from the survey), but only 26% of all outpatients and only 11% of those not needing hospital treatment (Table 2, see also Fig. 1). It is assumed that all victims who, according to the police, were taken to hospital and either were not admitted or whose admission is not known, were treated by the hospital as outpatients. To complete the picture, when compared with death certificate data, the police data on road accident deaths have not been shown to be less than 100%. Their coverage, therefore, declines rapidly with injury severity. This is in agreement with the findings of Tunbridge and Everest (1988). If the police were to record every casualty where they had been present, their data would still not be representative for injury severity. It would contain 83% of all inpatients, 48% of all outpatients, and 35% of all those not treated in a hospital (Table 3). (In addition, the completeness of data for Material Damage Only accidents is probably considerably lower than the 11% of those not treated in hospital.) Very much underrepresented were cyclists (11%) (Table 4, see also Fig. 2), especially, children on bikes (5%), and single vehicle bicycle accidents (2%). Single vehicle accidents in general (in accordance with Tunbridge and Everest 1988), and especially those in which not even an object was hit, so-called “single-sided accidents” in Dutch, (5%) were much underrepresented (9%), (Table 5, see also Fig. 3). The same applied to children in general (9%). All these results reflect also the incompleteness of police data in comparison with hospital inpatient data (Maas and Harris 1984). Relatively overrepresented, i.e. considerably more than the average of 24%, were car occupants (41%), especially those between 18 and 29 years old (54%). This also applied to accidents between two vehicles of which one was a car (40%) or one a moped (54%) and to accidents occurring at night between the hours of 10 P.M. and 7 A.M. This last result was unexpected because we had thought at night, with less police on duty and more crime, reporting would be less complete than during the daytime. The largest category of road user as the number of injured are concerned are not car occupants, which is what the police data indicates, but cyclists-they make up half Table 4. Completeness Mode of transport
n
Car Bicycle Moped Motorcycle Pedestrian Rest
50 122 41 6 18 3
Total
240
Casualties
by mode of transport Police
Completeness
Factor
44,000 106,000 36,000 5,0@3 16,000 3,000
18,269 11,734 11,998 1,961 4,076 1,710
41% 11% 33% 39% 25% 57%
2.4 9.0 3.0 2.5 3.9 1.8
210,000
49,748
24%
4.2
(xl = 100.86; df = 5; P < 0.01)
S.HARRIS
376
Table 5. Completeness
by collision type
Collision type
n
Casualties
Police
Completeness
Factor
with: Vehicle Object Nothing else Unknown
130 30 77 3
114,000 26,000 67,000 -
40,055 5,523 3,258 912
35% 21% 5% -
2.9 8.0 21 -
Total
240
210,000
49,748
24%
4.2
(x’ = 252.83; df = 3; P < 0.01)
of all wounded; not twens (people in their twenties), but children up to the age of fourteen; no mopedists between 15 and 29 years old, but child cyclists; not cars in collision with another vehicle, but single-vehicle cycle accidents. The ranking in the casualty rate (per milliard kilometers travelled) by category of road user does not alter much if the survey results are used instead of the police data. As far as age group is concerned the ranking does change; according to the survey data, children up to the age of 14 years become much more important in comparison with the police data. Women, not men, have the higher casualty rate. The early-evening and lunch hours become much more important, and the late-evening and nighttime hours less so. The ratios between weekdays and weekend, and inside and outside built-up areas change little. The two factors of seemingly greatest importance for determining whether an injury accident will be recorded are: injury severity and involvement of a motor vehicle. RECOMMENDATIONS
Following are recommendations road traffic accidents:
for improving
the recording
of those injured in
1. The public, those involved in the accident as well as bystanders, should report more injury accidents to the police. 2. The police should report all injury accidents conforming to the definition instead of only half of those where they were present. The expectations are low as far as improving the police recording in the future is concerned. Further computerization at the scene of accidents and at police stations, to decrease the amount of time needed to record each accident, does not guarantee the recording of more accidents. This is confirmed by certain police authorities here. Extra manpower for accident recording is certainly not readily available and may never be. There are greater hopes for the already intensively used integral inpatient data of the National Medical Registration as complementary to police data. Since 1987 all hospitals participate in this registration. 100 j
%
100
80
70% irr 60 P. a I 40 _ i e c 20 - n t 8
0
I
0
-
80
-
60
-
40
-
20
26% 11%
out-patients
notinhospital I
40
I
I
80
I
I
I
120
Fig. 1. Completeness
I
160
by treatment.
I
I'
200
0 x1000
Real number of road traffic accident casualties
60 40 20 0 210
0 Fig. 2. Completeness
by mode of transport.
3. The main recommendation is to obtain more nearly complete outpatient data by adding road traffic accident victims to the existing continuous Home and Leisure Accident Surveillance System of the Consumer Safety Institute; a representative 10% sample of 14 hospital Accident and Emergency Departments. Since the original draft report was presented to the Ministry of Transport it has commissioned SWOV to carry out a trial registration. This ran in two hospitals (one a teaching hospital, the other a general hospital) during the months of September, October, and November 1989. If successful (i.e. the necessary details can be obtained), the addition of road accident victims will become continuous. 4. Further recommended is exploration of the possibility of recording road victims by general practitioners, using the existing continuous General Practitioner Panel; a panel of 60 General Practitioners whose 45 practices cover 1% of the population. 5. Including road accident victims in these two registrations (outpatients and General Practitional patients), in combination with the existing inpatient data, would result in a continuous recording of practically all those receiving professional medical treatment (95% according to this survey). 6. The possible establishment of a separate Road Accident Recording Corps is also recommended. This would complement the police rather in the same way as Parking Wardens do. 7. Also recommended is the possibility of making this present one-year survey continuous, or at least repeating it, for example, every five years. %
%
100
100 80 -
-
80
60 -
-
60
35%
-
40
vehide
-
20
40 -
..,. 20 -
0
.’ , 0
I 40
I
‘. I
I
80
I
120
I
I
160
I
I 200
0 x1000
NB: the width of each staff is proportional to the number of survey injured. Fig. 3. Completeness
by collision type.
378
S. HARRIS
It must, however, be emphasized that the additional registrations recommended are primarily of victims and not accidents. They do not, therefore, contain, and probably never will do, many of the precrash and crash variables available in the police data. There is, however, the theoretical possibility of follow-up interviews with patients. These would have to take place soon after the accident and there are considerable privacy problems in getting patients’ names and addresses. As long as this is the case the police data remain indispensable for road safety research. INTERNATrONAL
IMPLrCATIONS
The extent of incompleteness of records, especially of outpatients, the extra-murally treated, and accidents not involving a motor vehicle, will not be the same in every country and will probably vary greatly. For example, the number of police-reported casualties for every reported fatality varied by more than a factor 4 between developed countries in 1987 (United Nations 1989)! In Denmark and Ireland there were only 17 casualties per fatality, the highest in Europe was 60 for Great Britain (in the Netherlands 33), and in Japan there were 77. It is, however, clear from these results for the Netherlands that the situation in other developed countries, whether better or worse than in the Netherlands, will also be cause for concern. In the United States the incompleteness of records is apparently much less- only 30% ~Greenblatt et al. 1981). Similar surveys in other countries will undoubtedly show to what extent reliance on the official (police) accident data can lead to inappropriate priorities. They will also show, therefore, that complementary data, as recommended by Hauer and Hakkert (1987) is needed. In our opinion, to a great extent this should be hospital data that include outpatients as well as inpatients. At the moment international comparisons of road safety should be limited to data on fatal accidents and fatalities. Use of police data on injury accidents and casualties should be avoided because of their great incompleteness and lack of representativity. They are not a valid indicator of road safety-in fact, they are misleading. Use of only those (nonfatal) victims being admitted to hospital (inpatients) may be valid if (a) the criteria for admission, and (b) the completeness of the police data are approximately the same in most countries; or at least known from year to year ~OE~~/SWOV 1988).
REFERENCES Greenblatt, J.; et al. National Accident Sampling System nonreported accident survey. Washington: National Highway Traffic Administration; 1981. Harris, S. Verkeersgewonden geteld en gemeten (Traffic casualties counted and measured)-with English summary. (R-89-13), Leidschendam: SWOV; 1989. Hauer, E.; Hakkert, A.S. The extent and some implications of incomplete accident reporting. 66th TRB meeting. Washington, DC: Transportation Research Board; 1987. Maas, M.W.; Harris, S. Police recording of road accident inpatients, Accid. Anal. Prev. 16: 167-184; 1984. Montfoort, G.L.M. van; Galen, dr. W.Ch.C. van; Harris, S. Ongevallen in Nederland (Accidents in the Netherlands)-with English summary. Amsterdam: Stichting Consument en Veiligheid; 1988. OECDISWOV. Framework for consistent traffic and accident statistical data bases. Leidschendam: SWOV; 1988. Tunbridge, R.J.; Everest, J.T. An assessment of the under reporting of road accident casualties in relation to injury severity. IRCOBI Conference, Bergisch Gladbach (FRG), September, 1988. United Nations. Statistics of road traffic accidents 1987, data and annex 1: Definitions and general notes. New York: UN; 1989. WHO. Manual of the international statistical classification of diseases, injuries and causes of death. Geneva: World Health Organization; 1977.